A design for smooth transition of robotic emotional states

Abstract

A two-dimensional (2-D) emotional model is proposed to represent the emotional state of a sociable robot. We present a novel design of autonomous robotic emotional state transition based on a fuzzy-neuro network. By using fuzzy Kohonen clustering network (FKCN), a smooth transition of the robotic emotional states is obtained to generate continuous emotional behaviors. Moreover, the robotic emotional character can be specified in this design by assigning weights to the FKCN. In this study, four emotional characters and five primitive emotional behaviors (boring, smiling, crying, angry and relaxed) have been designed for a robotic face. The method has been tested by a graphical simulator to verify the robotic responses to the user's emotional intensity (neutral, happiness, anger and sadness). Computer simulations show that the proposed emotional state transition scheme effectively responds to user's emotional intensity in a continuous manner. Practical experiments on a 16-DOF robotic face validate the proposed design.